Research article Special Issues

Inference and other aspects for qWeibull distribution via generalized order statistics with applications to medical datasets

  • Received: 03 January 2024 Revised: 27 January 2024 Accepted: 02 February 2024 Published: 27 February 2024
  • MSC : 62B10, 62G30

  • This work utilizes generalized order statistics (GOSs) to study the q-Weibull distribution from several statistical perspectives. First, we explain how to obtain the maximum likelihood estimates (MLEs) and utilize Bayesian techniques to estimate the parameters of the model. The Fisher information matrix (FIM) required for asymptotic confidence intervals (CIs) is generated by obtaining explicit expressions. A Monte Carlo simulation study is conducted to compare the performances of these estimates based on type Ⅱ censored samples. Two well-established measures of information are presented, namely extropy and weighted extropy. In this context, the order statistics (OSs) and sequential OSs (SOSs) for these two measures are studied based on this distribution. A bivariate q-Weibull distribution based on the Farlie-Gumbel-Morgenstern (FGM) family and its relevant concomitants are studied. Finally, two concrete instances of medical real data are ultimately provided.

    Citation: M. Nagy, H. M. Barakat, M. A. Alawady, I. A. Husseiny, A. F. Alrasheedi, T. S. Taher, A. H. Mansi, M. O. Mohamed. Inference and other aspects for qWeibull distribution via generalized order statistics with applications to medical datasets[J]. AIMS Mathematics, 2024, 9(4): 8311-8338. doi: 10.3934/math.2024404

    Related Papers:

    [1] Zhuonan Wu, Zengtai Gong . Algebraic structure of some complex intuitionistic fuzzy subgroups and their homomorphism. AIMS Mathematics, 2025, 10(2): 4067-4091. doi: 10.3934/math.2025189
    [2] Doaa Al-Sharoa . (α1, 2, β1, 2)-complex intuitionistic fuzzy subgroups and its algebraic structure. AIMS Mathematics, 2023, 8(4): 8082-8116. doi: 10.3934/math.2023409
    [3] Muhammad Jawad, Niat Nigar, Sarka Hoskova-Mayerova, Bijan Davvaz, Muhammad Haris Mateen . Fundamental theorems of group isomorphism under the framework of complex intuitionistic fuzzy set. AIMS Mathematics, 2025, 10(1): 1900-1920. doi: 10.3934/math.2025088
    [4] Aneeza Imtiaz, Hanan Alolaiyan, Umer Shuaib, Abdul Razaq, Jia-Bao Liu . Applications of conjunctive complex fuzzy subgroups to Sylow theory. AIMS Mathematics, 2024, 9(1): 38-54. doi: 10.3934/math.2024003
    [5] Aneeza Imtiaz, Umer Shuaib . On conjunctive complex fuzzification of Lagrange's theorem of ξ−CFSG. AIMS Mathematics, 2023, 8(8): 18881-18897. doi: 10.3934/math.2023961
    [6] Adeel Farooq, Musawwar Hussain, Muhammad Yousaf, Ahmad N. Al-Kenani . A new algorithm to compute fuzzy subgroups of a finite group. AIMS Mathematics, 2023, 8(9): 20802-20814. doi: 10.3934/math.20231060
    [7] Wen Li, Deyi Li, Yuqiang Feng, Du Zou . Existence and stability of fuzzy Pareto-Nash equilibria for fuzzy constrained multi-objective games with fuzzy payoffs. AIMS Mathematics, 2023, 8(7): 15907-15931. doi: 10.3934/math.2023812
    [8] Supriya Bhunia, Ganesh Ghorai, Qin Xin . On the fuzzification of Lagrange's theorem in (α,β)-Pythagorean fuzzy environment. AIMS Mathematics, 2021, 6(9): 9290-9308. doi: 10.3934/math.2021540
    [9] Supriya Bhunia, Ganesh Ghorai, Qin Xin . On the characterization of Pythagorean fuzzy subgroups. AIMS Mathematics, 2021, 6(1): 962-978. doi: 10.3934/math.2021058
    [10] Tareq M. Al-shami, Salem Saleh, Alaa M. Abd El-latif, Abdelwaheb Mhemdi . Novel categories of spaces in the frame of fuzzy soft topologies. AIMS Mathematics, 2024, 9(3): 6305-6320. doi: 10.3934/math.2024307
  • This work utilizes generalized order statistics (GOSs) to study the q-Weibull distribution from several statistical perspectives. First, we explain how to obtain the maximum likelihood estimates (MLEs) and utilize Bayesian techniques to estimate the parameters of the model. The Fisher information matrix (FIM) required for asymptotic confidence intervals (CIs) is generated by obtaining explicit expressions. A Monte Carlo simulation study is conducted to compare the performances of these estimates based on type Ⅱ censored samples. Two well-established measures of information are presented, namely extropy and weighted extropy. In this context, the order statistics (OSs) and sequential OSs (SOSs) for these two measures are studied based on this distribution. A bivariate q-Weibull distribution based on the Farlie-Gumbel-Morgenstern (FGM) family and its relevant concomitants are studied. Finally, two concrete instances of medical real data are ultimately provided.



    The Pareto system of distributions includes four types (I–IV) of cumulative distribution functions that produce different density shapes, see Johnson et al. [1]. The type II Pareto distribution is known in the literature by the Lomax distribution, see Lomax [2]. Lomax distribution is considered a serious distribution of lifetime models because it belongs to the family of decreasing failure rate models. When the population has a heavy-tailed distribution and the experimenter seeks a good probability distribution describing this case, the Lomax model is an excellent alternative to most common models such as the exponential, Weibull, or gamma distributions, see Bryson [3]. This distribution could be a flexible model because it's used for analyzing different lifetime data. It has various applications in medicine, biological sciences, engineering, operations research, natural sciences, queuing theory, and internet traffic modeling. One can see, among many others, Johnson et al. [1]. Recently, David et al. [4] found many applications for this distribution in business, economics, and actuarial modelling. Additionally, it's noted that this distribution has some characteristics that make it a preferred distribution. One in every of these, the statistical and reliability properties of this distribution may be expressed in closed forms, which help statisticians model different types of data.

    Due to these features, several authors used this model in modeling data, especially, censored observations. Hence our motivation to verify this distribution throughout this paper is because of its practical importance in the many different alternative areas mentioned above, as mentioned in many references. Among these references, Harris [5] and Atkinson and Harrison [6] found this distribution very convenient when examining data on income and wealth. Dubey [7] reported that this distribution is one of the special cases of the specific compound gamma distribution. Bryson [3] suggested that when the available data are heavy-tailed, the Lomax distribution is another suitable distribution to the exponential distribution. Tadikamalla [8] listed that the Lomax distribution can be derived from a family of Burr distributions. In addition to any or all of the above, this distribution has itself been used as a basis for several generalizations, see Al-Awadhi and Ghitany [9], Ghitany et al. [10], and Punathumparambath [11].

    If X is a continuous random variable that follows a Lomax distribution with shape and scale parameters θ and β, respectively, then the probability density function (PDF) and its corresponding cumulative distribution function (CDF) can be formulated as

    f(x;θ,β)=θβθ(β+x)(θ+1); x>0,

    and

    F(x;θ,β)=1βθ(β+x)θ; x>0,

    respectively. Further, the Lomax failure rate function can be expressed as

    h(x;θ,β)=θ(β+x)1; x>0,

    where β>0 and θ>0. In industrial operations, under normal operating conditions, we need a very long time to obtain the failure times of these products, which also causes the average failure times of these products to be long, this is not in line with the industrial revolution and modern technology. To provide quick solutions in such cases, experimenters perform the ALT, where the units we want to test, are placed under several stress levels whose values are higher than the normal stress values in order to reduce the time it takes to fail. Under typical conditions, the experimenters use the data obtained from the accelerated test to estimate the failure distribution of the units.

    The types of stress loading in ALT are generally classified as constant stress, step stress, or progressive stress. The constant-stress loading is a time-independent test setting where the stress remains fixed until an item is desert. It has different advantages when compared to time-dependent stress loads. This is simply, because most real products operate under constant stress. There are several major references in the area of ALT, among the most important of these references, the follower can look at Nelson [12], Meeker and Escobar [13], Bagdonavicius and Nikulin [14], among others.

    ALT first assumption is that there is a known or assumed relationship between life stresses, therefore, data generated from accelerated conditions can be relied upon for normal use conditions. However, it has been found that in some cases this relationship cannot be known or even adequately assumed, especially when the units under test are new. Therefore, partially accelerated life tests (PALT) are oftentimes used in such cases. In general, PALT can be divided into two types, the first is constant-stress PALT and the other is step-stress PALT. In constant-stress PALT, all groups of test units are placed separately in use conditions and accelerated conditions. For step-stress PALT, the test conditions for the remaining items in the experiment shift from conditions of use to conditions of higher stress at a given time or when a specified number of failures occur.

    Constant-stress PALT is our area of interest in this paper. In this type, all test units are divided into two groups, where the first group is assigned to work under normal conditions while the second group is assigned to work under accelerated conditions. In recent years, a lot of studies have been proposed in the field of PALT. For instance, we can refer to Ismail [15], Bhattacharyya and Soejoeti [16], Gouno [17], El-Morshedy et al. [18], Nassar et al. [19], Rao [20], Bai and Chung [21], Hassan and Al-Ghamdi [22], Wu and Kus [23], Wu et al. [24], Fan et al. [25], and Lio and Tsai [26]. Although the main objective of PALT is to shorten the test time of the experiment, the experimenter loses a lot of time waiting for all test units to fail. For this, it is necessary to deal with censored data. In literature, type-I and type-II censoring schemes are the main censoring schemes. These do not permit the withdrawal of intermediate components. Because of this constraint, these schemes are not flexible enough. To avoid the difficulties of these schemes, various generalizations have been proposed. One of these is the progressive censoring scheme. It allows for live units to be removed from the test at time points (controlled conditions) other than the final termination point of the experiment. The article presented by Balakrishnan and Aggarwala [27] has many applications in progressive type-II censoring. Sometimes, some problems arise concerning the age of the products, and therefore the experimental time of control is simply too long. To resolve this issue, the statistician provided other censoring schemes. Among them was Johnson [28], who introduced the scheme of the first failure censoring.

    During this censoring, the test units whose number is N=n×k are distributed over the n number of groups where each group has k number of test units, after that, a life test is performed for all groups at the same time and conditions and the first failure times are recorded in each of the test groups. With this life-telling design, a great deal of time and cost is saved. Thus, it is clear that both first failure and first failure progressive censoring schemes significantly improve the efficiency of the life test. From this point, Wu and Kus [23] presented the progressive first-failure censoring scheme which profits the practical advantages of the two schemes. There are several advantages that distinguish this censoring, including the test time is short as well as the resources are saved, and a few units at risk will be removed during the experiment at each of several ordered failure times. Another advantage of this censoring is that it allows the experimenters to remove one or more of the groups under test even without noticing the first failure in the groups that will be removed from the experiment. Hence, this censoring is more efficient in reliability studies. Recently, more literature is available on progressive first-failure censoring, including Lio and Tsai [26], Soliman et al. [29], Ahmed [30], Krishna et al. [31], Kumar et al. [32], El-Din et al. [33], Ahmadi and Doostparast [34], and Saini et al. [35].

    Now, we briefly discuss the first-failure progressive type-II censoring scheme. Let n independent groups with k items in each group are put on a life test. Start dropping R1 number of groups as soon as the first failure time X1:m:n:k has occurred. When the second failure time X2:m:n:k occurs, randomly delete R2 groups from the experiment and also delete the group in which the second failure was observed and repeat. The experimenter continues in the same manner until all remaining live Rm groups and the group in which the mth failure has occurred are removed at the time of the mth failure. The observed failures X1:m:n:k<X2:m:n:k<...<Xm:m:n:k called progressive first failure censored order statistics and while =(R1,...,Rm) is called a progressive censoring scheme. This scheme includes various kinds of censoring schemes. For example, if k=1, it reduces to a progressive censoring scheme. Also, it contains the usual order statistics (k=1, =(0,0,...,0) and m =n) and type II right censoring scheme (k=1,R1,...,Rm1=0 and Rm=nm) as special cases.

    Before completing this section, a brief presentation on the parameter estimation of the Lomax distribution should be made. There are several studies presented by many authors focusing on estimating the parameters of the Lomax distribution when observations are completed or censored. The Bayesian and classical estimators for the same sample size from Lomax distribution based on the progressive type-I censoring, are discussed by Elfattah et al. [36]. Raqab et al. [37] presented different predictors of failure times based on multistage progressive censoring from Pareto distribution. The optimal censoring scheme for estimating the parameters of the Lomax distribution based on progressive type-II censoring is presented by Cramer and Schmiedt [38]. Empirical Bayes estimators of reliability performances using progressive type-II censoring from the Lomax model are discussed by El-Din et al. [33]. Al-Zahrani and Al-Sobhi [39] have concerned with the estimation problem of the P(Y<X) of the Lomax distribution based on general progressive censoring. When the available data is progressive type-II censoring data, the MLEs of the Lomax distribution are derived using the expectation-maximization (EM) algorithm by Helu et al. [40]. Wei et al. [41] studied Bayes estimation of the Lomax distribution parameters in the composite LINEX loss of symmetry. Furthermore, based on the progressively type-I hybrid censored sample, Asl et al. [42] applied the classical and Bayesian inferential procedures for the Lomax distribution. Chandra and Khan [43] have derived a step-stress ALTs model for the Lomax failure time under modified progressive type-I censoring. Recently, under generalized progressive hybrid censoring, Mahto et al. [44] analyzed a partially observed competing risk model for the Lomax model.

    Meanwhile, no articles appeared on estimating parameters of the Lomax distribution based on a constant-stress PALT under progressive first failure type-II censored data. Our aim in this article is to address this research point. Two approaches will be used to estimate the parameters of the Lomax distribution, the maximum likelihood (ML) and Bayesian methods. When calculating the maximum likelihood estimates (MLEs) and their associated confidence intervals, the Newton-Raphson (NR) and EM algorithms, are derived and discussed in detail. On the other hand, when the Bayes procedure is performed, there is difficulty in obtaining the posterior distributions because this process often requires the computation of integrals, which is often difficult to compute, whether using complex high or low-dimensional models. In such a case, we can rely on the MCMC to estimate the unknown parameters. Assuming independent gamma priors for the parameters, the Bayes estimates will be performed under different loss functions. Based on various techniques of the MCMC, the Gibbs within Metropolis samples can be applied to generate samples from the posterior distributions. The performance of the Bayes estimators can be compared with classical MLEs through extensive computer studies. Approximate 95% confidence intervals of the unknown parameters can be calculated based on the MLEs via NR, EM, and MCMC algorithms. Also, we will compare them through their average lengths and coverage probability.

    As for the form and organization of this paper as well as its sequence, it is as follows: Model description and basic assumptions are presented in Section 2. Frequentest estimations including MLEs and approximate confidence intervals of unknown parameters, are provided in Section 3. During this section, we also developed the EM algorithm for calculating the MLEs and the observed Fisher information matrix. Bayes point estimates via (symmetric-asymmetric) loss functions and their corresponding credible intervals, are constructed in Section 4. A real-life example as well as simulation studies are presented and discussed in Section 5. Finally, a few concluding remarks are reported in Section 6.

    As explained above, in constant-stress PALT, N1 number of test units are randomly selected from all available units (N), and are operated under normal conditions, while the remaining N2=NN1 number of units are operated under accelerated conditions. Assume the lifetime of the test unit follows the Lomax distribution, therefore, the PDF, CDF, and failure rate function at normal conditions are given, respectively by

    f1(x1,θ,β)=θβθ(β+x1)(θ+1); x1>0, (1)
    F1(x1;θ,β)=1βθ(β+x1)θ ; x1>0, (2)

    and

    h1(x1;θ,β)=θ(β+x1)1; x1>0, (3)

    where β>0 and θ>0. The hazard rate function for any item tested at the accelerated condition is given as follows

    h2(x2;θ,β,λ)=λh1(x1)=θλ(β+x2)1; x2>0, (4)

    where λ (λ>1) is the acceleration factor. Under the accelerated condition, the PDF and its corresponding CDF are obtained, respectively by

    f2(x2;θ,β,λ)=θλβθλ(β+x2)(θλ+1); x2>0, (5)

    and

    F2(x2;θ,β,λ)=1βθλ(β+x2)θλ; x2>0. (6)

    Here, the progressive first-failure censoring is combined with constant-stress PALT. Hence, the total of N test units will be divided into two groups, namely, group I and group II. Elements of the first group (N1=n1k1) are assigned to normal conditions whereas the items of the second group (N2=n2k2) are allocated to stress conditions. Each group is divided under normal or accelerated conditions into several groups containing the number of test units kd. In this scheme, R1i and R2i are the progressive censoring schemes for normal and accelerated tests, respectively. This scheme runs until md; d=1,2 failures are observed in each test condition, respectively, as follows.

    For group I (items dedicated to normal condition) or group II (items are dedicated to stress condition), suppose that a random sample Xd=(Xd1,Xd2,...,Xndkd) of size Nd=ndkd,; d=1,2 is put on a life test experiment where the lifetimes Xdis are independent and identically distributed (IID) random variables with PDF fXd(xd;Θ) and CDF FXd(xd;Θ) where Θ is a vector of unknown parameters, and Nd test units are divided into nd independent groups with kd items in each group. Now, when the first failure Xd1 is appeared, then the group with observed failure and also Rd1 of the remaining (nd1) groups are randomly removed from the experiment. Next, when the second failure Xd2 is observed, then the group with the second observation and also Rd2 of the remaining (ndRd11) groups are randomly removed. The experiment continues until the mdth failure with lifetime Xdmd is observed and the group with this observed failure and the remaining Rdmd groups are all removed from the experiment. In this way, a progressive first-failure censored sample Xd=(Xd1,Xd2,...,Xdmd) of size md can be obtained from a parent sample of size Nd with censoring scheme d=(Rd1,Rd2,...,Rdmd) where mdi=1Rdi=ndmd. In our study, Rdi,i=1,2,...,md are fixed prior and md<nd. With a progressively first-failure type-II censoring scheme under constant-stress PALT, the likelihood function of the observed sample Xd can be written as

    L(x;Θ)2d=1{mdi=1fd(xdi;Θ)[1Fd(xd(i);Θ)]kd(Rdi+1)1}. (7)

    Regarding the proposed PALT approach, the following assumptions are made:

    (A1) The total number of units under test is N=n1k1+n2k2.

    (A2) The lifetime of all units tested under various normal or accelerated conditions follow the Lomax distribution.

    (A3) The lifetimes of test items are IID random variables.

    (A4) The lifetimes X1i, i=1,2,,m1 of items assigned to the normal condition, while the lifetimes X2i, i=1,2,,m2 of items assigned to the accelerated condition are mutually independent.

    (A5) The lifetime of any item at accelerated condition is X2=λ1X1.

    Next based on a progressive first failure censored sample from the Lomax distribution, the MLEs via both NR and EM algorithms are derived for the unknown parameters θ, β, and λ.

    Consistency, asymptotic efficiency, asymptotic unbiased, and asymptotic normality are some of the characteristics of the MLE, which made it one of the most important and popular methods for fitting the statistical model. In this section, the MLEs of θ, β, and λ can be derived from the likelihood function (LF) presented in (7). The substitution of (1)–(6) into (7) yields

    L(x;Θ)2d=1mdi=1θSd(λ)βθSd(λ)(β+xdi)(θSd(λ)+1)[βθSd(λ)(β+xdi)θSd(λ)]kd(Rdi+1)1θm1+m2βθ(k1n1+λk2n2)λm22d=1mdi=1(β+xdi){θSd(λ)kd(Rdi+1)+1}, (8)

    where

    Sd(λ)={1,  d=1λ,  d=2. (9)

    The MLEs of θ, β and λ are the values that maximize the LF in (8). Maximizing the LF is difficult, so it is preferable to maximize the logarithm of this function, where the log-LF, say =logL(θ,β,λ|y), can be formulated as

    (θ,β,λ)(m1+m2)logθ+m2logλ+θ(k1n1+λk2n2)logβ2d=1mdi=1(θSd(λ)kd(Rdi+1)+1)log(β+xdi). (10)

    By calculating the first derivatives of (10) with respect to θ, β, and λ and then setting them equal to zero, the resulting simultaneous equations are expressed as follows

    (θ,β,λ)θ=(m1+m2)θ+(k1n1+λk2n2)logβ2d=1mdi=1[Sd(λ)kd(Rdi+1)]log(β+xdi)=0, (11)
    (θ,β,λ)β=θ(k1n1+λk2n2)β2d=1mdi=1[θSd(λ)kd(Rdi+1)+1]β+xdi=0,                (12)

    and

    (θ,β,λ)λ=m2λ+θk2n2logβθk2m2i=1(R2i+1)log(β+x2i)=0.                           (13)

    The preceding equations mathematically represent a system of three nonlinear equations in three unknowns θ, β,  and λ. Theoretically, there is difficulty in providing closed-form solutions to the previous nonlinear equations. Thus, the numerical NR technique will be applied to solve these simultaneous equations to get the MLEs (ˆθML, ˆβML, ˆλML) of (θ, β, λ).

    Theorem 1. The MLEs of the parameters θ>0, β>0 and λ>1 are exist and unique.

    Proof. To prove this theorem, three cases should be discussed, namely, (i) for given β>0 and λ>1, the MLE of θ exists and is unique; (ii) for given θ>0 and λ>1, the MLE of β exists and is unique; and (iii) for given θ>0 and β>0, the MLE of λ exists and is unique.

    For the first case, when β and λ are known, the MLE of θ can be obtained from (11) as follows

    ˆθML=(m1+m2)/ϕ(β,λ),

    where

    ϕ(β,λ)={2d=1mdi=1[Sd(λ)kd(Rdi+1)]log(β+xdi)(k1n1+λk2n2)log(β)}.

    Using (10) and ˆθML, we can write

    (θ,β,λ)(m1+m2)logθ+ω1(θ,β,λ)=(m1+m2)logθ(m1+m2)logˆθML+(m1+m2)logˆθML+ω(θ,β,λ)=(m1+m2)log(θˆθML)+(m1+m2)logˆθML+ω(θ,β,λ),

    where

    ω(θ,β,λ)=m2logλ+θ(k1n1+λk2n2)log(β)2d=1mdi=1(θSd(λ)kd(Rdi+1)+1)log(β+xdi).

    Using inequality logθˆθMLθˆθML1, we have

    (θ,β,λ)(m1+m2)θˆθML(m1+m2)+(m1+m2)logˆθML+ω(θ,β,λ)θ|ˆθML.

    Hence,

    (θ,β,λ)θϕ(β,λ)(m1+m2)+(m1+m2)logˆθ+ω(θ,β,λ)θ|ˆθML.

    Equality holds if and only if θ=ˆθML or equivalent, (θ,β,λ)=(ˆθML,β,λ) if and only if θ=ˆθML. This proves that the function (θ,β,λ) reaches its maximum value at the point (ˆθ,β,λ) where β and λ are given. Using the similarity steps to prove (ii) and (iii).

    For more results about the existence and uniqueness theorem of MLEs of the Lomax parameters, see Cramer and Schmiedt [38], and recently Qin and Gui [45]. The second derivatives are very important for distributions that are expressed using over one parameter for several reasons. For one of those reasons, they are going to confirm that maxima have been identified. In our case, the second partial derivatives of the LF can be expressed as

    {2θ2=(m1+m2)θ2,θβ=βθ=(k1n1+λk2n2)β2d=1mdi=1Sd(λ)kd(Rdi+1)(β+xdi),θλ=λθ=k2n2log(β)k2m2i=1(R2i+1)log(β+x2i),             2β2=θ(k1n1+λk2n2)β2+2d=1mdi=1[θSd(λ)kd(Rdi+1)+1](β+xdi)2,2λ2=m2λ2,2βλ=θk2n2βθk2m2i=1(R2i+1)β+x2i. (14)

    Arranging the second partial derivatives (14) in matrix form yields the Hessian matrix, which when multiplied by -1 in a statistical context, is known as a Fisher information matrix. In an optimization context, a sufficient condition for a stationary point to be a maximum is that the Hessian matrix is negative semi-definite. In the statistical literature, the inverse form of the Fisher information matrix is noted by the variance-covariance matrix. Thus, for the MLEs of the Lomax parameters, the Hessian matrix can be proposed as

    H=[2θ22θβ2θλ2βθ2β22βλ2λθ2λλ2λ2].

    The corresponding Fisher information matrix is

    I=[2θ22θβ2θλ2βθ2β22βλ2λθ2λβ2λ2],

    and the asymptotic variance-covariance matrix is

    I1=[Var(ˆθML)Cov(ˆθML,ˆβML)Cov(ˆθML,ˆλML)Cov(ˆβML,ˆθML)Var(ˆβML)Cov(ˆβML,ˆλML)Cov(ˆλML,ˆθML)Cov(ˆλML,ˆβML)Var(ˆλML)]. (15)

    Thus, an approximate (1α)100% confidence intervals for θ, β, and λ are obtained from the asymptotic normality of the ML results of intervals calculated according to

    ˆθMLzα/2Var(ˆθML), ˆβMLzα/2Var(ˆβML), ˆλMLzα/2Var(ˆλML), (16)

    where zα/2 is defined as the percentile of the standard normal model with α/2 right-tail probability. In the next two sections, the EM algorithm is applied to calculate MLEs of unknown parameters and their corresponding confidence intervals.

    In this segment, our focus will be on the application of the EM algorithm in estimating the Lomax parameters, as it represents a general algorithm to obtain the MLEs, especially, in the problems of incomplete or missing data. The standard NR algorithm does not converge in some cases (see Pradhan and Kundu [46]), while the convergence is reliably an advantage of the EM method over other methods. Thus, the EM algorithm is utilized in this subsection as a good alternative to the NR approach, in order to calculate the MLEs of the unknown parameters of the Lomax model based on the fact that the available sample (constant partially PALT under progressive first-failure type-II censored samples) is considered incomplete data. This algorithm was proposed by Dempster et al. [47] and later was thoroughly discussed alongside its extensions in the book by McLachlan and Krishnan [48]. Now, let us symbolize the observed data by Xd=(Xd1,Xd2,...,Xdmd) whereas the censored data by Zd=(Zd1,Zd2,...,Zdmd), where each Zdi is 1×(kdRdi+kd1) vector with Zdi=(Zdi1,Zdi2,...,Zdi(kdRdi+kd1)); i=1,2,...,md; d=1,2 and they are not observable. Let W=(Xd,Zd) represents the complete data set, hence the complete data LF will be in the following form

    Lc(W;Θ|R)=2d=1{mdi=1fd(xdi;θ,β,λ)mdi=1kd(Rdi+1)1j=1fd(zdij;θ,β,λ)}. (17)

    After ignoring the constants in the previous equation, the log-LF can be written as

    c(W;Θ)=Nln(θ)+θ(k1n1+k2λn2)lnβ+n2k2ln(λ)2d=1mdi=1(Sd(λ)θ+1)ln(β+xdi)2d=1mdi=1kd(Rdi+1)1j=1(Sd(λ)θ+1)ln(β+Zdij). (18)

    The EM algorithm runs in two main stages: The expectation stage (E-stage), followed by the maximization stage (M-stage), which are repeated until achieving at least one convergence criterion. At each iteration, the missing data are filled with expected data and consequently, the estimations of the parameters are updated. About the requirements of the E-stage, one needs to compute the pseudo-log-LF, which can be obtained from c(W;θ,β,λ) by replacing any function of zdij, say g(zdij), with an accompanying conditional expectation "E[g(zdij)|zdij>Xdi]". Therefore,

    Ls(Θ)=E(c(W;Θ)|Xd)=Nln(θ)+(k1n1+k2λn2)θlnβ+n2k2ln(λ)2d=1mdi=1(Sd(λ)θ+1)ln(β+xdi)2d=1mdi=1kd(Rdi+1)1j=1(Sd(λ)θ+1)E[ln(β+Zdij)]. (19)

    For simplicity, xdi has been used instead of xdi:md:nd:kd; d=1,2. Then, the M-stage begins, in which the pseudo-log-LF in (19) is maximized relative to θ, β, and λ. Thus, at sth stage, the estimate of (θ,β,λ) is (θ(s),β(s),λ(s)), then (θ(s+1),β(s+1),λ(s+1)) can be obtained by maximizing Ls(θ,β,λ) with respect to (θ,β,λ). Hence, by equating the partial derivative of Ls(Θ) to zero, a nonlinear system of equations can be generated as follows

    (k1n1+k2n2)(1θ+lnβ)2d=1mdi=1Sd(λ)ln(β+xdi)2d=1mdi=1(kd(Rdi+1)1)Sd(λ)A(xdi;θ,β,λ)=0, (20)
    θ(k1n1+λk2n2)β2d=1mdi=1(Sd(λ)θ+1)β+xdi2d=1mdi=1(kd(Rdi+1)1)(Sd(λ)θ+1)B(xdi;θ,β,λ)=0, (21)

    and

    n2k2λ+θn2k2ln(β)θm2i=1log(β+x2i)θm2i=1(k2(R2i+1)1)A(x2i;θ,β,λ)=0.     (22)

    The conditional distribution of zdi follows the truncated Lomax distribution with left truncation at xdi (see Ng et al. [49]). That is,

    f(zdi|zdi>xdi)=f(zdi)1F(xdi); zdi>xdi, i=1,2,...,md, d=1,2.

    Hence,

    f(zdij|zdij>xdi)=Sd(λ)θ(β+xdi)Sd(λ)θ[β+zdij](Sd(λ)θ+1); zdij>xdi, d=1,2. (23)

    Then, the conditional expectation E(ln(β+zdi)|zdi>xdi) can be obtained as follows

    E(ln(β+zdi)|zdi>xdi)=Sd(λ)θ(β+xdi)Sd(λ)θxdiln(β+t)[β+t](Sd(λ)θ+1)dt,

    using u=ln(β+t) and dv=(β+t)(Sd(λ)θ+1)dt,

    A(xdi;Θ)=E(ln(β+zdi)|zdij>xdi)=ln(β+xdi)+1Sd(λ)θ, (24)

    and

    B(xdi;Θ)=E[(β+zdi)1|zdi>xdi]=Sd(λ)θ(β+xdi)Sd(λ)θxdi(β+zdi)(Sd(λ)θ+2)dt,=Sd(λ)θ(Sd(λ)θ+1)(β+xdi).

    Depending on the method presented by Louis [50], we can have an observed Fisher information matrix. Let X indicate observed data, W denotes the complete data, IX(Θ) shows the observed information, while IW(Θ) and IW|X(Θ) stand for the complete and missing information, then according to the Louis approach we have

    IX(Θ)=IW(Θ)|Θ=(θ,β,λ)IW|X(Θ)|Θ=(θ,β,λ),,Θ=(θ,β,λ)), (25)

    where IW(Θ) is obtained as

    IW(Θ)|Θ=(θ,β,λ)=E[fW(w;Θ)Θ2]Θ=(θ,β,λ)=[a11a12a13a21a22a23a31a32a33]. (26)

    The complete data LF of the Lomax distribution in constant-stress PALT can be expressed as

    fW(w;Θ)=2d=1ndkdi=1fd(xd(i);Θ)=θnλn2k2βθ(k1n1+k2n2λ)2d=1ndkdi=1(β+xdi)(Sd(λ)θ+1). (27)

    Hence, the log-LF based on Xd is given by

    lnfW(w;Θ)=nln(θ)+θ(k1n1+k2n2λ)ln(β)+n2k2ln(λ)2d=1ndkdi=1(Sd(λ)θ+1)log(β+xdi), (28)

    hence,

    {a11=E[2lnfW(w;Θ)θ2]=nθ2,a33=E[2lnfW(w;Θ)λ2]=n2k2λ2,a22=E[2lnfW(w;Θ)β2]=θ(k1n1+k2n2λ)β22d=1ndkdi=1(Sd(λ)θ+1)E(β+xdi)2,a12=a21=E[2lnfW(w;Θ)βθ]=(k1n1+k2n2λ)β1+2d=1ndkdi=1Sd(λ)E[(β+xdi)1],a13=a31=E[2lnfW(w;Θ)θλ]=k2n2ln(β)+n2kdi=1E[log(β+x2i)],a23=a32=E[2lnfW(w;Θ)βλ]=k2n2θβ1+θn2k2i=1E[(β+x2i)1],

    where

    E[(β+xdi)1]=Sd(λ)θβSd(λ)θ01(β+t)(Sd(λ)θ+2)dt=Sd(λ)θβ(Sd(λ)θ+1),   (29)
    E[(β+xdi)2]=Sd(λ)θβSd(λ)θ0dt(β+t)(Sd(λ)θ+3)=Sd(λ)θβ2(Sd(λ)θ+2), (30)

    and

    E[log(β+xdi)]=Sd(λ)θβSd(λ)θ0log(β+t)(β+t)(Sd(λ)θ+1)dt=1+Sd(λ)θlog[β]Sd(λ)θ.   (31)

    The Fisher information matrix for a single observation which censored at ith time failure is given by

    IiW|X(Θ)=E[2lnfzdi(zdi|xdi,Θ)Θ2]Θ=(θ,β,λ)=[b11(xdi,Θ)b12(xdi,Θ)b13(xdi,Θ)b21(xdi,Θ)b22(xdi,Θ)b23(xdi,Θ)b31(xdi,Θ)b32(xdi,Θ)b33(xdi,Θ)], (32)

    where fzdi(zdi|xdi,Θ) is given in (23) and the expected values of the second partial of the log-LF of Z given X are calculated as

    {b11(xdi,Θ)=E[2lnfzdi(zdi|xdi,Θ)θ2]=1θ2,b22(xdi,Θ)=E[2lnfzdi(zdi|xdi,Θ)β2]=θSd(λ)(2+θSd(λ))(β+xdi)2,b33(xdi,Θ)=E[2lnfzdi(zdi|xdi,Θ)λ2]=(Sd(λ)Sd(λ))2,b12(xdi,Θ)=b21(xdi,Θ)=E[2lnfzdi(zdi|xdi,Θ)θβ]=Sd(λ)(β+xdi)(Sd(λ)θ+1),b13(xdi,Θ)=b31(xdi,Θ)=E[2lnfzdi(zdi|xdi,Θ)θλ]=Sd(λ)θSd(λ),b23(xdi,Θ)=b32(xdi,Θ)=E[2lnfzdi(zdi|xdi,Θ)βλ]=θSd(λ)(1+θSd(λ))(β+xdi),

    consequently, the total missing information is obtained as

    IX(Θ)|Θ=(θ,β,λ)=2d=1mdi=1[kd(Rdi+1)1]IiW|X(Θ)|Θ=(θ,β,λ), (33)

    where IiW|X(Θ)|Θ=(θ,β,λ) is given by (35). Finally, the asymptotic variance-covariance matrix of the MLE of Θ is then obtained as

    [IX(Θ)]1Θ=(θ,β,λ)=E[2lnL(xd;Θ)Θ2]1Θ=(θ,β,λ). (34)

    Once the relevant variance estimates have been obtained from MLEs for θ, β, and λ, asymptotic confidence intervals of 100(1γ)% can be easily generated using the normality property of MLEs.

    In this section, the Bayes estimates for the unknown parameters θ, β, and λ are derived and discussed. Here, we consider independent gamma priors for θ and β, with PDFs given, respectively by

    πθ=θa11eb1θ; a1>0, b1>0, (35)

    and

    πβ=βa21eb2β; a2>0, b2>0, (36)

    where (a1,b1) and (a2,b2) are hyper-parameters chosen to reflect prior knowledge about θ and β. The non-informative distribution with ai=0,  bi=0; i=1,2 can be chosen if no prior knowledge is available. Moreover, a vague prior is selected for the acceleration factor λ with the following PDF

    πλ=λ1; λ>0. (37)

    The joint prior density function of θ, β, and λ is then obtained by

    π(θ,β,λ)=θa11βa21λ1eb1θb2βb3λ. (38)

    The joint posterior distribution of θ, β, and\ λ is reported using Bayes theorem by combining the LF of (θ,β,λ) given in (8) and the joint prior distribution π(θ,β,λ) in (42) as

    π(θ,β,λ)=Aθ(m1+m2+a11)βθ(k1n1+k2n2λ)+a21λm2+a31eb1θb2βb3λ×2d=1mdi=1(1+βxdi)[Sd(λ)θkd(Rdi+1)+1]=Aθ(m1+m2+a11)βθ(k1n1+k2n2λ)+a21λm21×exp[b1θb2β2d=1mdi=1[Sd(λ)θkd(Rdi+1)+1]ln(β+xdi)], (39)

    where A is a normalizing constant and is given by

    A=001π(θ,β,λ)dθdβdλ.

    In Bayes' approach, to reach the best estimator, one should choose a loss function corresponding to each of the possible estimators. Here, the estimates are derived concerning two different types of loss functions, namely, symmetric loss functions and asymmetric loss functions. Squared error loss function (SELF) will be a representative of the first type, while both general entropy loss function (GELF) and LINEX loss function (LLF) will represent examples of choosing the second type. When an overestimation or underestimation occurs, the SELF is not appropriate. In this case, GELF and LLF can be taken as an alternative selection to estimate the parameters. The LLF is useful when overestimation is more serious than underestimation and vice-versa. Let ψ be an estimator for the unknown parameter ϕ, the loss functions under study can be presented as follows:

    ● Symmetric loss function, SELF: Form "(ψϕ)2" and Bayes estimate "Eϕ(ϕ|t_)".

    ● Asymmetric loss function:

    - GELF: Form "(ψϕ)qqln(ψϕ)1,  q0" and Bayes estimate "(Eϕ(ϕq|t_))1q".

    - LLF: Form "exp[c(ψϕ)]c(ψϕ)1, c0" and Bayes estimate "1cln(Eϕ(exp(cϕ)|t_))".

    It is not difficult to notice that, when q=1, the Bayes estimation for any parameter ϕ based on GELF and SELF are the same (see Calabria and Pulcini [51]). Referring to (43), we can notice the difficulty of calculating integrals, and then the inability to obtain the joint posterior in a closed form that enables us to calculate Bayes estimations of the unknown parameters θ, β, and λ. Therefore, in order to obtain these estimates, we will rely on the MCMC approach, which enables us to obtain simulated samples from the posterior distributions of the parameters. These generated samples will be used for calculating the point and interval estimation of unknown parameters. As for the mechanism of this method, it depends on the calculation of conditional posterior functions where the conditional distribution of θ given β and\ λ can be expressed as

    π1(θ|β,λ,x_)θ(m1+m2+a11)exp[θT(x;Sd(λ),β)]Gamma[m1+m2+a1,b1+T(x;Sd(λ),β)], (40)

    where

    T(x;Sd(λ),β)=2d=1mdi=1(Sd(λ)kd(Rdi+1))ln(β+xdi)(k1n1+λk2n2)ln(β). (41)

    Similarly, the conditional distribution of β given θ, λ, and data can be reported as

    π2(β|θ,λ,x_)βθ(k1n1+k2n2λ)+a21exp[b2β2d=1mdi=1(Sd(λ)θkd(Rdi+1)+1)ln(β+xdi)]. (42)

    Further, the conditional distribution of λ given θ, β, and data, can be listed as

    π3(λ|θ,β,x_)λm21exp[λ{m2i=1θk2(R2i+1)ln(β+x2i)θk2n2ln(β)}]Gamma[m2,m2i=1θk2(R2i+1)ln(β+x2i)θk2n2ln(β)]. (43)

    It can be seen that π1(θ|β,λ,x_) and π3(λ|θ,β,x_) are gamma densities. Thus, samples of θ and λ can be generated using a gamma generator. Moreover, π2(β|θ,λ,x_) cannot be reduced for drawing samples directly by standard methods. In such a case, to obtain Bayes' estimate for β, we can use one of the well-known algorithms in the MCMC method, which is Metropolis-Hastings (MH) algorithm model, which was presented in the literature by Metropolis et al. [52]. To apply this algorithm, we need to assume a proposal function to sample from it. In this algorithm, we may choose either symmetric or non-symmetric proposal distribution to decrease the rejection rate as much as possible. Since the marginal distribution of β is not well-known, then the normal distribution is listed as a symmetric proposal distribution.

    The Metropolis-Hastings steps are included in the Gibbs sampler to update β, while θ as well as λ is updated directly from its full conditional. Below is a hybrid algorithm with Gibbs sampling steps for updating the parameters θ and λ with MH steps for updating β:

    Step 1: Choose an initial guess of (θ,β,λ), say (θ(0),β(0),λ(0)), and set l=1.

    Step 2:Generate β(l) according to the following steps:

    (i) Generate β from normal N(β(l1),Var(ˆβML)) distribution where Var(ˆβML) denotes the variance of β;

    (ii) Compute r=min{1,π2(β|θ(l1),λ(l1),x_)π2(β(l1)|θ(l1),λ(l1),x_)};

    (iii) Generate a sample u from the U(0;1) distribution;

    (iv) If ur, then set β(l)=β; otherwise β(l)=β(l1).

    Step 3:Generate θ(l) from Gamma(m1+m2+a1,m1+m2+a1,b1+T(xd;Sd(λ(l1)),β)).

    Step 4:Generate λ(l) from Gamma(m2,m2i=1θ(l)k2(R2i+1)ln(β(l)+x2i)θ(l)k2n2ln(β(l))).

    Step 5:Set l=l+1.

    Step 6:Repeat steps (25) M times to obtain the desired number of samples.

    We discard the initial M0 number of burn-in samples and obtain estimates using the remaining MM0 samples. Hence, the Bayes estimate of ϕ=(θ,β, or λ) under SELF can be considered as the mean of the generated samples from the posterior densities as follows

    ˆϕBS1MM0Ml=1ϕl. (44)

    Further, the approximate Bayes estimate for ϕ, under LLF and GELF are then given, respectively by

    ˆϕBL=1clog[1MM0Ml=M0+1ecϕ(l)], (45)

    and

    ˆϕBGE=[1MM0Ml=M0+1(ϕq)(l)]1q. (46)

    To construct the highest posterior density (HPD) credible interval of ϕ=(θ, β, or λ) using generated MCMC sampling procedure, we first refer to the ordered random sample generated in the previous algorithm in the form ϕ(1)<ϕ(2)<...<ϕ(M), then the 100(1γ)% confidence intervals for the parameter ϕ can be listed as

    (ˆϕ(lM),ˆϕ(l+[M(1γ)]M)); l=1,2,...,M[M(1γ)], (47)

    where [.] refers to the greatest integer function.

    In this section, we introduce a numerical investigation of the estimation methods discussed in previous sections for the Lomax distribution using simulated data and a realistic data set.

    Here, the estimation procedure described in the previous sections is applied to the set of simulated progressive first-failure type-II censoring sample data under the constant-stress PALT. A data set of system lifetime is generated from the Lomax model with θ=0.2, β=2 and λ=1.2, respectively. Based on N=200 (n1=n2=50, k1=k2=2), using the algorithm described in Balakrishnan and Sandhu [53], we simulate two samples of size m1=20 and m2=30 from the Lomax(θ,β) and Lomax(θλ,β), respectively under the following two progressive censoring schemes

    1={5,0,0,5,0,0,3,0,0,0,5,2,2,2,1,1,1,1,1,1},2={3,0,0,0,2,0,0,0,2,0,0,0,2,0,0,0,2,0,0,3,0,1,0,1,0,1,0,0,2,1}.

    The simulated data are given in Table 1.

    Table 1.  Simulated progressive first-failure censored samples with constant PALT.
    Normal condition 0.0600 0.1763 0.2126 0.4183 0.4304 0.5200 0.7656
    1.8601 1.9396 2.2424 3.0073 3.4157 5.7828 7.2197
    8.2609 23.9444 49.6628 61.0012 185.615 235.921
    Accelerated condition 0.0080 0.1242 0.3281 0.4551 0.7045 0.8303 1.0725
    1.2998 1.3810 1.7342 1.8328 2.2544 2.3445 4.1533
    5.0055 5.1700 5.8743 6.0516 6.6369 6.7635 7.2544
    7.7786 8.1127 14.6889 27.7101 28.4127 33.5706 74.577
    151.143 218.760

     | Show Table
    DownLoad: CSV

    For the generated data, we first compute the MLEs of θ, β, and λ using the NR and EM methods, the results are listed in Table 2. Of the tabulated values, the estimates obtained using the NR method are far more distant from the true parameter value than those obtained under the EM algorithm. Then, next, we begin to calculate Bayes estimates, depending on the proposed loss functions and using informative priors (a1=20, b1=100, a2=200, b2=100). For LLF, c=10 and 10 were used as two options for the constant c. These choices give more weight to underestimation and exaggeration, respectively. Similarly, two choices for q such as q=10 and 10 are considered under GELF. The results are reported also in Table 2. From Table 2, it can be seen that, the estimates under the SELF, LINEX, and the GELF are close to the actual values of the parameters. From (15) and (34), the observed variance-covariance matrix based on NR and EM\ algorithms are given, respectively by

    (0.002212280.0300040.01003150.0300040.9033390.01429830.01003150.01429830.140199),

    and

    (0.001807240.01363110.0108840.01363110.3573740.01429830.0108840.01429830.138958).

    Table 2 also contains all interval estimations of the unknown parameters θ, β, and λ using the various methods mentioned. The Bayesian estimator gives narrower credible intervals as compared to the classical MLEs.

    Table 2.  Estimated values of θ, β and λ and its corresponding 95% confidence intervals (CI).
    Method MLEs Bayes estimators
    Parameter NR EM SELF LLF GELF
    c1=10 c2=10 c1=10 c2=10
    θ=0.2 Estimate 0.1559 0.1605 0.1775 0.1816 0.1737 0.1973 0.1535
    95% CI (0.0637,0.2481) (0.0771,0.2438) (0.1262,0.2349)
    β=2 Estimate 1.9077 1.9886 2.0040 2.1091 1.9093 2.0482 1.9486
    95% CI (0.0449,3.7705) (0.8169,3.1603) (1.7395,2.2833)
    λ=1.2 Estimate 1.2970 1.2887 1.1914 2.0059 0.9254 0.1973 0.1535
    95% CI (0.5632,2.0309) (0.5580,2.0192) (0.7188,1.8596)

     | Show Table
    DownLoad: CSV

    The numerical example shows that the means of the Bayes estimates and MLEs via NR or EM algorithm of the unknown parameters θ, β and λ are close to the true values. It is seen that the Bayes estimates have better performances than MLEs where the Bayes estimates feature smaller bias than MLEs and that the interval lengths of HPD credible intervals are shorter than approximate confidence intervals, which also shows that HPD credible intervals are superior to approximate confidence intervals in terms of interval lengths.

    In this segment, we would like to know how the estimation techniques suggested in the above sections work in practice for the accelerated data set. The data are from the accelerated life test of oil breakdown times of insulating fluid under various levels of high voltage. Table 3 shows this data under two high test voltages stress levels (in kilovolt, Kv), see Nelson [12]. Here, the stress level (32 Kv) was assumed as normal conditions while the other stress level (36 Kv) was considered to be the accelerated conditions.

    Table 3.  Times to breakdown of an insulating fluid.
    Normal conditions 0.27 0.40 0.69 0.79 2.75 3.91 9.88 13.95
    (32 Kv) 15.93 27.80 53.24 82.85 89.29 100.58 215.10
    Accelerated conditions 0.35 0.59 0.96 0.99 1.69 1.97 2.07 2.58
    (36 Kv) 2.71 2.90 3.67 3.99 5.35 13.77 25.50

     | Show Table
    DownLoad: CSV

    The initial shape is reported using the non-parametric approaches like histogram, kernel density, box, violin, TTT, and normal Q-Q plots in Figures 1 and 2. It is noted that the data is asymmetric as well as some outliers observations were founded. Moreover, the failure rate is decreasing.

    Figure 1.  Non-parametric plots for normal conditions.
    Figure 2.  Non-parametric plots for accelerated conditions.

    Before further proceeding, we first check if the Lomax model can be used as a proper one to fit this data through the Kolmogorov-Smirnov (K-S) statistic. K-S distances and its associated p-values were calculated for the data set with the two stress levels, and the calculated values were: 0.1671 (0.7964) and 0.1435 (0.9169), respectively. Thus, we ensured that the Lomax distribution is considered an appropriate distribution for this data. The MLEs are listed in Table 4. By moving to Bayes estimates, since no prior information is available about the unknown population parameters, the noninformative (or vague) gamma priors are appropriate in this case. In this case the hyperparameters take zero values (ai=bi=0, i=1,2). As noted earlier, Metropolis within the Gibbs algorithm was relied upon to generate 10000 MCMC samples based on the MLEs of θ, β, and λ as initial values at the beginning of the algorithm. Trace plots for the first 10000 MCMC outputs of θ, β, and λ are shown in Figure 3. It is showing good convergence of MCMC procedure. Moreover, the histogram plots of generated samples of θ, β, and λ were presented in Figure 4. It is observed that the histograms of the generated samples quite well with the theoretical posterior density functions.

    Figure 3.  Trace plots of the parameters based on the MCMC outputs for Nelson data.
    Figure 4.  Histograms of the parameters obtained from the Gibbs sampling based on Nelson data.
    Table 4.  Estimated values of θ, β and λ and its corresponding 95% CI for complete Nelson data.
    Method MLEs Bayes estimators
    Parameter NR EM SELF LLF GELF
    c=2 c=2 q=5 q=5
    θ Estimate 0.7726 0.7758 0.9191 0.3579 0.3451 0.3603 0.3243
    95% CI (0.110242,1.43492) (0.166638,1.37853) (0.552597,1.38861)
    β Estimate 6.9467 6.9568 9.742 11.1155 8.8626 9.9055 9.6694
    95% CI (4.27344,18.1667) (2.60354,16.4968) (7.80366,11.97200)
    λ Estimate 3.1321 3.1333 3.5563 6.2863 2.3876 3.289 2.6017
    95% CI (0.449858,5.81441) (0.603516,5.66076) (1.66313,6.67387)

     | Show Table
    DownLoad: CSV

    For Bayesian estimations under asymmetric LLF, it is known in the literature that, c<0 implies that underestimation results in more penalty than overestimation and the reverse is true for c>0. When c close to zero LLF becomes symmetric, and it behaves roughly as the SELF. The Bayes estimates of θ, β, and λ are computed and reported in Table 4 with c=2 and 2. The second asymmetric loss used here is GELF. In this loss function q>0 means overestimation is more serious than underestimation and the opposite is true when q<0. The symmetrical case in which this loss function approximately corresponds to SELF occurs when q=1. Also, the Bayes estimates are computed and recorded in Table 4 with q=2 and 2. This table also displays the 95% ML and Bayes confidence intervals with the corresponding lengths for all estimates based on the complete sample data.

    Go ahead, based on the complete failure data in Table 3, in order to reduce cost and time, we assume that the number of items put on a life test is equal to nd×kd, d=1,2, where nd denotes the number of groups and kd the number of items in each group. The previous data were randomly grouped into n1=n2=5 sets, with k1=k2=3 observations in each. Here, It is assumed that we observed only m1=m2=4 data with arbitrarily chosen censoring schemes 1=2=(1,0,0,0). Thus, the progressive first failure censored samples under the two stress are obtained as follows: X1={0.27,2.75,3.91,82.85} and X2={0.35,1.69,2.07,2.71}. The MLEs relative to both NR, EM techniques and Bayesian MCMC method of unknown parameters θ, β, and λ are computed and listed in Table 5 for Nelson data under schema V: n1=n2=5, k1=k2=3, m1=m2=3. Moreover, the results of 95% approximate CI, credible intervals, and HPD intervals of θ, β, and λ are given in Table 5.

    Table 5.  Estimated values of θ, β and λ and its corresponding 95% CI for Nelson data under schema V.
    Method MLEs Bayes estimators
    Parameter NR EM SELF LLF GELF
    c=2 c=2 q=2 q=2
    θ Estimate 0.3354 0.3358 0.6344 0.6911 0.5892 0.6727 0.5157
    95% CI (0.110242,1.43492) (0.166638,1.37853) (0.552597,1.15338)
    β Estimate 5.0338 5.0438 9.74342 10.8765 8.8918 9.7954 9.5889
    95% CI (4.27344,18.1667) (2.60354,16.4968) (7.93259,11.8236)
    λ Estimate 3.315 3.3169 3.5606 22.5007 1.8117 4.3008 1.8284
    95% CI (0.449858,5.81441) (0.603516,5.66076) (0.75323,9.76393)

     | Show Table
    DownLoad: CSV

    Finally, we can conclude that the estimated Lomax distribution provides an excellent good fit for the given data and the Bayes estimates fits the data better than MLEs. Also, we have decreased the number of units used here and yet we have obtained results close to the complete sample.

    In this numerical section, we compare the performance of various estimators and confidence intervals mentioned above by conducting a simulated study. First, under normal and accelerated conditions from the Lomax(θ, β) and Lomax(θλ, β) distributions, respectively, the algorithm which is given by Balakrishnan and Sandhu [53] can be utilized to generate progressive first-failure censored samples. This simulation was earned by considering different values of (n1,n2,m1,m2,k1,k2), and by choosing (θ,β,λ)=(0.4,1.5,1.3) in all cases. Further, we are giving results only for two different group sizes (k1,k2)=(1,2) and (2,1),  and the censoring schemes used in this simulation are:

    Scheme I:R1d,...,R2d,...,R(md1)d=0,Rmd=ndmd,d=1,2.Scheme II:{R1d=R2d=...=Rmd12=0;Rmd+12=ndmd,if md is odd. R1d=R2d=...=Rmd22=0;Rmd2=ndmd,if md is is even.Scheme III:R2d,...,R3d,...,Rmdd=0,R1d=ndmd,d=1,2.Scheme IV:R1d,...,R3d,...,Rmdd=0,nd=md,d=1,2.

    In order to improve the appearance of the numerical tables, the various censoring schemes have been represented by short notations as (10,04) denotes (10,0,0,0,0). In each case, estimates of θ, β, and λ were calculated based on both the ML and Bayes methods. The NR and EM algorithm methods were used for ML estimation computation, while the MH algorithm with the Gibbs sampling algorithm is used to compute Bayes estimators. The informative prior distributions for both θ and β, with hyperparameters a1, b1, a2, and b2 were utilized and the associated hyperparameters are defined through the following discussion. According to the idea of Arabi Belaghi et al. [54], let M1 random samples from Lomax(θ=0.4,β=1.5) distribution are available, and that (ˆθi,ˆβi); i=1,2,,M1 are the MLEs of (θ, β). Suppose that a parameter ϕ is a prior distributed as gamma with density proportional to ϕh11eh2ϕ, where ϕ can be either θ or β. Then prior mean and variance are obtained by h1h2and h1h22, respectively. Now, equating the sample mean and associated variance of ˆϕi with the mean and variance of the prior distribution, we get

    1M1M1i=1ˆϕi=h1h2 and 1M11M1i=1[ˆϕiM1j=1ˆϕj]2=h1h22.

    From these equations, we get

    h1=[1M1M1i=1ˆϕi]21M11M1i=1[ˆϕiM1j=1ˆϕj]2 and h2=[1M1M1i=1ˆϕi]1M11M1i=1[ˆϕiM1j=1ˆϕj]2.

    Thus, if the unknown parameter is θ, then the corresponding hyperparameters are estimated as h1=a1,h2=b1 and likewise, for β, we have h1=a2 and h2=b2. Following this, we compute hyperparameters by taking M1=1000 and for each M1 a sample of size 50 is taken into consideration to obtain desired estimates. Consequently, hyperparameters are assigned values as a1=2.021, b1=4.9488, a2=5.827 and b2=3.2019. Considering different loss functions, all the Bayes estimators are derived, and the corresponding approximate estimates are provided by applying Gibbs sampling under the MH sampling technique (see Section 4) approach. The different values of loss parameters are arbitrarily taken as c=±1.5 and q=±1.5, which produces estimators under both symmetric and asymmetric loss functions. The MH with the Gibbs algorithm is implemented using the normal distribution as the proposal distribution. Markov chains of size 5000 are generated and the first 1000 of the observations are removed to eliminate the effect of the starting distribution. While utilizing this algorithm, we take the corresponding MLEs as an initial guess for the respective unknown parameters and the associated variance-covariance matrix as the variance-covariance matrix of the normal proposal density. All the simulations were performed using the Mathematica 9 software and the results were based on NS=1000 Monte Carlo runs. We compute average values (Avg) and corresponding mean square error (MSE) of different estimates. Here,

    Avg =NSiˆϕi/NS  and MSE={ˆϕiϕ}2/NS,

    where ˆϕ is the estimate of ϕ. Also, we obtain the average length (AL) of 95% confidence/HPD credible intervals and 95% CP of the parameters based on the simulation. The results of the Monte Carlo simulation study are presented in Tables A1–A4 (see appendix). From these tables the following conclusions are made:

    (1) The MSEs of all estimates decrease as the sample sizes increase in all cases, as expected.

    (2) The MLEs of θ, β, and λ using the EM algorithm have smaller MSEs than the MLEs using the NR algorithm. Hence, the MLEs via the EM algorithm perform better than those obtained by the NR method.

    (3) In most simulations, the Bayes estimates outperform the MLEs for the estimation of θ, β, and λ. So, in general, we would recommend using the Bayes estimate of the unknown parameters of Lomax distribution based on a first-failure progressively censoring scheme under constant-stress PALT. Also, from the tabulated values, we notice that the Bayes estimates of θ, β, and λ under asymmetric loss function (LLF, GELF) are sensitive to the value of the parameters c and q. It is observed that, as expected, the positive values of c and q respectively under LINEX and GELF lead to smaller estimates as compared to the negative value of c and q. Moreover, the choice c=1.5 for the LINEX loss seems to be reasonable. For the GELF, q=1.5 is a better value in computing the Bayes estimate based on the LINEX or the SELE. Hence, it can be concluded that the asymmetric loss function makes the Bayes estimates attractive for use, in reality, the scale parameters c and q of the LINEX and GELF make one estimate the unknown parameters with more flexibility.

    (4) Finally, in most cases, the MSEs and the AL of the intervals for (θ,β,λ) are maximized when the censoring is at the end. Hence, when we compare progressive type-II censoring and type-II censoring plans, the progressive type-II plan provides a better result.

    In this work, different parameter estimation methods for the two parameters of the Lomax distribution are discussed based on constant stress partially accelerated life test using the progressive first failure type-II censored sample data. We derived the MLEs and Bayes estimates of the parameters as well as the acceleration factor and the corresponding confidence intervals. We have also exploited the EM algorithm to obtain the MLEs for the unknown parameters. In addition, the supplemented EM algorithm was used for finding the asymptotic covariance matrix of the ML estimators. Meanwhile, because the explicit expressions of MLEs of part of parameters cannot be derived, we look to the Bayesian method for support. Based on the square error, LINEX, and general entropy loss functions, the Bayes estimates are obtained on the premise of independent gamma priors. The posterior distributions of unknown parameters indicated that some parameters do not follow a well-known distribution. Consequently, we utilized MH sampling within the Gibbs sampling steps algorithm to calculate the Bayes estimates with associated HPD credible intervals. In order to compare the performance of all mentioned methods directly and to investigate the effect of different values of n, m, and k, the simulation study was implemented. The average, MSEs, the average widths of the interval estimators, and the coverage percentages of each estimator are illustrated. The simulation results show that for fixed values of m and k, the performances of both estimation methods improve with n, and for fixed n and k, their performances improve with m, also for fixed n and m, their performances improve with k.

    Further, the numerical results demonstrated that the Bayes method based on informative priors outperformed the ML method under both NR and EM methods. Also, it is perceived that the estimates obtained under the EM method generally perform well compared to those obtained under the NR method in terms of yielding relatively low values of MSEs and average widths of the interval estimators. Further, the Bayes estimates based on LINEX, and general entropy loss functions are more efficient than the squared error loss function under informative and non-informative priors. Practically, the study has demonstrated that the Lomax model has provided good flexibility for modeling oil breakdown times of insulating fluid data sets. The novelty in this research is that if a progressively first failure type-II censored sample is used, different sample sizes (k1k2) can be considered in each group, and this is fully consistent with real examples when doing life tests. Finally, in this work, although we have mainly considered progressively first failure type-II censoring and Lomax distribution, the same method can be extended for other distribution and censoring schemes also. There are numerous further works to be done here, for example, the design of optimal censoring schemes, inference of competing risks model with more failure factors, and making statistical prediction of the further order statistics based on the PALTs from Lomax distribution, these topics can be investigated in the future. Finally, we recommend the use of MCMC, and EM procedures based on partially accelerated life test using the progressive first failure type-II censored on data related to life testing, reliability modeling as well as biological analysis.

    The authors extend their appreciation to the Deputyship for Research & Innovation, Ministry of Education, Saudi Arabia for funding this research work through the project number (QU-IF-2-5-3-25108). The authors also thank to Qassim University for technical support.

    The authors declare no conflicts of interest.

    Table A1.  Estimated values of θ, β and λ and its corresponding 95% CI for Nelson data under schema V.
    (k1,k2) n1=n2 m1=m2 Scheme MLE MCMC
    NR EM SE LLF GELF
    c=1.5 c=1.5 q=1.5 q=1.5
    (1,2) 40 35 I Avg 0.4437 0.4425 0.4322 0.4370 0.4275 0.4357 0.4148
    MSE 0.0149 0.0142 0.0066 0.0073 0.0061 0.0070 0.0053
    II Avg 0.4413 0.4395 0.4312 0.4362 0.4264 0.4349 0.4131
    MSE 0.0122 0.0115 0.0059 0.0066 0.0054 0.0063 0.0046
    III Avg 0.4488 0.4490 0.4314 0.4367 0.4263 0.4352 0.4124
    MSE 0.0209 0.0211 0.0067 0.0074 0.0062 0.0071 0.0054
    40 IV Avg 0.4347 0.4341 0.4272 0.4314 0.4231 0.4303 0.4117
    MSE 0.0100 0.0097 0.0059 0.0064 0.0055 0.0062 0.0049
    60 40 I Avg 0.4349 0.4311 0.4280 0.4321 0.4240 0.4310 0.4130
    MSE 0.0116 0.0100 0.0067 0.0072 0.0062 0.0069 0.0052
    II Avg 0.4319 0.4285 0.4254 0.4301 0.4209 0.4288 0.4083
    MSE 0.0114 0.0099 0.0061 0.0067 0.0057 0.0064 0.0050
    III Avg 0.4607 0.4456 0.4328 0.4387 0.4272 0.4370 0.4122
    MSE 0.0265 0.0182 0.0071 0.0079 0.0063 0.0075 0.0053
    50 I Avg 0.4209 0.4200 0.4182 0.4214 0.4151 0.4207 0.4060
    MSE 0.0070 0.0067 0.0042 0.0045 0.0040 0.0043 0.0036
    II Avg 0.4219 0.4202 0.4193 0.4228 0.4159 0.4220 0.4061
    MSE 0.0073 0.0067 0.0044 0.0047 0.0042 0.0045 0.0038
    III Avg 0.4326 0.4295 0.4255 0.4294 0.4216 0.4284 0.4110
    MSE 0.0098 0.0086 0.0049 0.0053 0.0045 0.0051 0.0039
    60 IV Avg 0.4216 0.4208 0.4195 0.4224 0.4168 0.4217 0.4088
    MSE 0.0057 0.0055 0.0037 0.0040 0.0036 0.0039 0.0032
    90 75 I Avg 0.4170 0.4156 0.4169 0.4191 0.4137 0.4186 0.4083
    MSE 0.0042 0.0038 0.0032 0.0034 0.0031 0.0033 0.0028
    II Avg 0.4181 0.4164 0.4182 0.4206 0.4158 0.4201 0.4089
    MSE 0.0039 0.0035 0.0029 0.0031 0.0028 0.0030 0.0025
    III Avg 0.4212 0.4196 0.4170 0.4196 0.4144 0.4190 0.4091
    MSE 0.0055 0.0047 0.0033 0.0035 0.0032 0.0034 0.0029
    90 IV Avg 0.4101 0.4096 0.4100 0.4118 0.4081 0.4114 0.4027
    MSE 0.0033 0.0031 0.0024 0.0026 0.0024 0.0025 0.0022
    (2, 1) 40 35 I Avg 0.4489 0.4458 0.4326 0.4387 0.4268 0.4369 0.4114
    MSE 0.0244 0.0215 0.0086 0.0096 0.0078 0.0091 0.0067
    II Avg 0.4512 0.4471 0.4338 0.4403 0.4277 0.4384 0.4113
    MSE 0.0194 0.0167 0.0079 0.0089 0.0071 0.0084 0.0059
    III Avg 0.4614 0.4551 0.4373 0.4444 0.4306 0.4422 0.4131
    MSE 0.0318 0.0276 0.0089 0.0101 0.0079 0.0095 0.0066
    40 IV Avg 0.4342 0.4314 0.4260 0.4313 0.4210 0.4299 0.4069
    MSE 0.0148 0.0130 0.0062 0.0068 0.0056 0.0065 0.0049
    60 40 I Avg 0.4432 0.4336 0.4328 0.4378 0.4279 0.4364 0.4145
    MSE 0.0165 0.0121 0.0071 0.0078 0.0065 0.0074 0.0056
    II Avg 0.4457 0.4344 0.4331 0.4392 0.4273 0.4374 0.4118
    MSE 0.0182 0.0130 0.0076 0.0085 0.0068 0.0081 0.0059
    III Avg 0.4522 0.4422 0.4384 0.4437 0.4333 0.4422 0.4197
    MSE 0.0165 0.0122 0.0076 0.0083 0.0068 0.0080 0.0058
    50 I Avg 0.4318 0.4308 0.4298 0.4342 0.4256 0.4330 0.4139
    MSE 0.0113 0.0105 0.0064 0.0069 0.0059 0.0067 0.0051
    II Avg 0.4230 0.4220 0.4200 0.4245 0.4156 0.4234 0.4033
    MSE 0.0100 0.0092 0.0051 0.0056 0.0047 0.0053 0.0042
    III Avg 0.4440 0.4365 0.4323 0.4377 0.4271 0.4362 0.4129
    MSE 0.0156 0.0118 0.0059 0.0067 0.0054 0.0063 0.0044
    60 IV Avg 0.4131 0.4120 0.4139 0.4174 0.4105 0.4166 0.4006
    MSE 0.0067 0.0055 0.0039 0.0042 0.0037 0.0041 0.0034
    90 75 I Avg 0.4166 0.4166 0.4183 0.4213 0.4154 0.4255 0.4072
    MSE 0.0066 0.0065 0.0041 0.0043 0.0039 0.0042 0.0034
    II Avg 0.4249 0.4202 0.4231 0.4264 0.4199 0.4256 0.4109
    MSE 0.0074 0.0055 0.0040 0.0043 0.0037 0.0041 0.0033
    III Avg 0.4349 0.4283 0.4284 0.4323 0.4246 0.4313 0.4142
    MSE 0.0124 0.0093 0.0049 0.0054 0.0045 0.0052 0.0039
    90 IV Avg 0.4170 0.4146 0.4193 0.4218 0.4168 0.4212 0.4096
    MSE 0.0042 0.0033 0.0030 0.0032 0.0029 0.0031 0.0026

     | Show Table
    DownLoad: CSV
    Table A2.  The Avg (first row) and MSEs (second row) of the ML and Bayes estimates for β when θ = 0.4, β=1.5, λ=1.3 under different schemes.
    (k1,k2) n1=n2 m1=m2 Scheme MLE MCMC
    NR EM SE LLF GELF
    c=1.5 c=1.5 q=1.5 q=1.5
    (1, 2) 40 35 I Avg 1.9487 1.9239 1.6463 1.8187 1.5017 1.6772 1.4853
    MSE 1.4301 1.3267 0.1386 0.2867 0.0811 0.1538 0.0963
    II Avg 1.8878 1.8641 1.6439 1.8298 1.4924 1.6768 1.4738
    MSE 1.0170 0.9452 0.1277 0.2827 0.0742 0.1430 0.0889
    III Avg 1.9244 1.9082 1.5971 1.7984 1.4394 1.6327 1.4156
    MSE 2.1821 1.9819 0.1277 0.2870 0.0832 0.1420 0.1003
    40 IV Avg 1.8533 1.8391 1.6184 1.7904 1.4791 1.6486 1.4627
    MSE 0.8758 0.8257 0.1292 0.2753 0.0790 0.1428 0.0833
    60 40 I Avg 1.8115 1.7579 1.6106 1.7351 1.4970 1.6341 1.4865
    MSE 0.7088 0.5538 0.1492 0.2520 0.0972 0.1605 0.1124
    II Avg 1.7342 1.6980 1.5992 1.7417 1.4766 1.6260 1.4617
    MSE 0.4861 0.3920 0.1160 0.2143 0.0788 0.1260 0.0916
    III Avg 1.9262 1.8033 1.6188 1.8203 1.4580 1.6546 1.4346
    MSE 1.3477 0.9358 0.1245 0.3002 0.0747 0.1405 0.0917
    50 I Avg 1.7261 1.7066 1.5934 1.7129 1.4880 1.6161 1.4760
    MSE 0.6082 0.5395 0.1249 0.2071 0.0882 0.1333 0.1014
    II Avg 1.7221 1.6990 1.6144 1.7545 1.4941 1.6402 1.4819
    MSE 0.5098 0.4512 0.1370 0.2452 0.0908 0.1480 0.1050
    III Avg 1.7670 1.7285 1.6204 1.7918 1.4825 1.6505 1.4676
    MSE 0.5894 0.4851 0.1271 0.2698 0.0778 0.1409 0.0904
    60 IV Avg 1.6947 1.6807 1.5949 1.7187 1.4871 1.6181 1.4758
    MSE 0.4309 0.3989 0.1229 0.2056 0.0862 0.1316 0.0805
    90 75 I Avg 1.6451 1.6238 1.5917 1.6821 1.5084 1.6094 1.5006
    MSE 0.2574 0.2084 0.1094 0.1637 0.0809 0.1154 0.0903
    II Avg 1.6388 1.6180 1.5948 1.6935 1.5060 1.6138 1.4974
    MSE 0.2187 0.1838 0.0957 0.1517 0.0689 0.1021 0.0767
    III Avg 1.6788 1.6577 1.5900 1.7052 1.4891 1.6119 1.4779
    MSE 0.3472 0.2775 0.1040 0.1756 0.0742 0.1115 0.0722
    90 IV Avg 1.6289 1.6202 1.5838 1.6746 1.5023 1.6012 1.4953
    MSE 0.2182 0.1962 0.0980 0.1498 0.0722 0.1039 0.0792
    (2.1) 40 35 I Avg 1.8044 1.7871 1.6128 1.7672 1.4842 1.3151 1.2090
    MSE 0.8652 0.7910 0.1230 0.2429 0.0788 0.1356 0.1064
    II Avg 1.8533 1.8339 1.6477 1.8209 1.5077 1.6778 1.4935
    MSE 0.8790 0.8127 0.1353 0.2794 0.0804 0.1499 0.0936
    III Avg 1.9007 1.8694 1.6678 1.8580 1.5162 1.7004 1.5011
    MSE 1.0838 0.9791 0.1371 0.3029 0.0762 0.1537 0.0897
    40 IV Avg 1.7283 1.716 1.6041 1.7598 1.4771 1.6323 1.4612
    MSE 0.5710 0.53798 0.1153 0.2241 0.0783 0.1257 0.0911
    60 40 I Avg 1.7212 1.6788 1.5865 1.6960 1.4867 1.6078 1.4746
    MSE 0.5303 0.4192 0.1036 0.1690 0.0743 0.1108 0.0841
    II Avg 1.7519 1.7062 1.6288 1.7706 1.5106 1.6541 1.4993
    MSE 0.4974 0.4075 0.1200 0.2317 0.0760 0.1314 0.0871
    III Avg 2.1016 1.9482 1.6413 1.8236 1.4921 1.6739 1.4736
    MSE 2.6694 2.0469 0.1309 0.2746 0.0803 0.1452 0.0863
    50 I Avg 1.6749 1.6685 1.6140 1.7324 1.5091 1.6363 1.4991
    MSE 0.3906 0.3709 0.1215 0.2052 0.0824 0.1306 0.0934
    II Avg 1.6700 1.6652 1.6063 1.7328 1.4971 1.6299 1.4851
    MSE 0.3408 0.3249 0.1042 0.1866 0.0705 0.1129 0.0804
    III Avg 1.7818 1.7451 1.6501 1.8033 1.5230 1.6771 1.5117
    MSE 0.6692 0.5746 0.1295 0.2567 0.0777 0.1428 0.0899
    60 IV Avg 1.6233 1.6184 1.5872 1.7000 1.4890 1.6085 1.4779
    MSE 0.2850 0.2621 0.1017 0.1671 0.0734 0.1085 0.0832
    90 75 I Avg 1.6283 1.6281 1.6134 1.7061 1.5305 1.6309 1.5247
    MSE 0.2590 0.2580 0.1129 0.1748 0.0809 0.1200 0.0893
    II Avg 1.6488 1.6286 1.6107 1.7040 1.5268 1.6285 1.5201
    MSE 0.2445 0.2086 0.0936 0.1454 0.0667 0.0999 0.0729
    III Avg 1.6911 1.6588 1.6166 1.7433 1.5272 1.6478 1.5195
    MSE 0.3804 0.3097 0.1131 0.1991 0.0744 0.1214 0.0832
    90 IV Avg 1.6164 1.6058 1.6120 1.6967 1.5338 1.6284 1.5278
    MSE 0.1747 0.1571 0.0925 0.1402 0.0659 0.0985 0.0716

     | Show Table
    DownLoad: CSV
    Table A3.  The Avg (first row) and MSEs (second row) of the ML and Bayes estimates for λ when θ = 0.4, β=1.5, λ=1.3 under different schemes.
    (k1,k2) n1=n2 m1=m2 Scheme MLE MCMC
    NR EM SE LLF GELF
    c=1.5 c=1.5 q=1.5 q=1.5
    (1.2) 40 35 I Avg 1.4152 1.4091 1.3222 1.4284 1.2438 1.3426 1.2245
    MSE 0.2032 0.1892 0.1211 0.1900 0.0961 0.1267 0.1067
    II Avg 1.3929 1.3868 1.3127 1.4164 1.2357 1.3330 1.2152
    MSE 0.1610 0.1495 0.0994 0.1589 0.0797 0.1040 0.0896
    III Avg 1.3759 1.3707 1.2918 1.3991 1.2148 1.3123 1.1939
    MSE 0.1845 0.1680 0.1146 0.1909 0.0932 0.1192 0.1054
    40 IV Avg 1.3803 1.3760 1.3047 1.3920 1.2375 1.3223 1.2198
    MSE 0.1618 0.1530 0.0987 0.1449 0.0821 0.1022 0.0907
    60 40 I Avg 1.3888 1.3744 1.3195 1.4070 1.2517 1.3371 1.2346
    MSE 0.1500 0.1297 0.1026 0.1516 0.0831 0.1068 0.0916
    II Avg 1.3740 1.3636 1.3158 1.4013 1.2485 1.3333 1.2310
    MSE 0.1290 0.1093 0.0940 0.1350 0.0776 0.0976 0.0855
    III Avg 1.3683 1.3464 1.2944 1.3836 1.2255 1.3126 1.2067
    MSE 0.1506 0.1186 0.0930 0.1355 0.0780 0.0963 0.0868
    50 I Avg 1.3570 1.3516 1.3083 1.3737 1.2541 1.3223 1.2400
    MSE 0.1115 0.0990 0.0804 0.1056 0.0689 0.0828 0.0743
    II Avg 1.3845 1.3766 1.3389 1.4096 1.2813 1.3536 1.2680
    MSE 0.1145 0.1014 0.0890 0.1233 0.0724 0.0926 0.0779
    III Avg 1.3677 1.3583 1.3158 1.3864 1.2585 1.3307 1.2439
    MSE 0.1164 0.1004 0.0778 0.1078 0.0649 0.0807 0.0705
    60 IV Avg 1.3509 1.3461 1.3121 1.3661 1.2664 1.3241 1.2542
    MSE 0.0695 0.0628 0.0553 0.0711 0.0484 0.0569 0.0516
    90 75 I Avg 1.3360 1.3304 1.3082 1.3498 1.2715 1.3177 1.2617
    MSE 0.0572 0.0493 0.0476 0.0572 0.0428 0.0487 0.0451
    II Avg 1.3461 1.3382 1.3194 1.3622 1.2816 1.3290 1.2721
    MSE 0.0620 0.0523 0.0530 0.0646 0.0467 0.0543 0.0489
    III Avg 1.3250 1.3210 1.2919 1.3343 1.2546 1.3017 1.2441
    MSE 0.0616 0.0516 0.0478 0.0567 0.0439 0.0486 0.0466
    90 IV Avg 1.3335 1.3306 1.3097 1.3440 1.2788 1.3177 1.2705
    MSE 0.0435 0.0383 0.0369 0.0434 0.0335 0.0376 0.0350
    (2,1) 40 35 I Avg 1.3239 1.3210 1.2969 1.3863 1.2280 1.3151 1.2090
    MSE 0.1339 0.1249 0.1123 0.1575 0.0957 0.1156 0.1064
    II Avg 1.3240 1.3229 1.3011 1.3916 1.2312 1.3196 1.2118
    MSE 0.1228 0.1157 0.1046 0.1500 0.0883 0.1079 0.0984
    III Avg 1.3097 1.3093 1.2838 1.3695 1.2167 1.3019 1.1965
    MSE 0.1060 0.0994 0.0877 0.1194 0.0779 0.0898 0.0870
    40 IV Avg 1.3254 1.3247 1.3020 1.3780 1.2411 1.3182 1.2241
    MSE 0.0963 0.0902 0.0827 0.1112 0.0719 0.0849 0.0791
    60 40 I Avg 1.3129 1.3190 1.2903 1.3662 1.2299 1.3064 1.2124
    MSE 0.0970 0.0864 0.0835 0.1126 0.0735 0.0855 0.0813
    II Avg 1.3243 1.3284 1.3007 1.3761 1.2403 1.3168 1.2230
    MSE 0.0869 0.0758 0.0749 0.1006 0.0659 0.0770 0.0725
    III Avg 1.4401 1.4304 1.3304 1.4400 1.2498 1.3514 1.2295
    MSE 0.2028 0.1856 0.1014 0.1650 0.0795 0.1070 0.0881
    50 I Avg 1.3228 1.3220 1.3018 1.3610 1.2522 1.3147 1.2387
    MSE 0.0858 0.0832 0.0763 0.0957 0.0677 0.0779 0.0728
    II Avg 1.3315 1.3305 1.3099 1.3688 1.2604 1.3229 1.2470
    MSE 0.0770 0.0744 0.0679 0.0859 0.0600 0.0696 0.0645
    III Avg 1.3083 1.3106 1.2901 1.3477 1.2417 1.3029 1.2276
    MSE 0.0745 0.0681 0.0668 0.0828 0.0607 0.0679 0.0659
    60 IV Avg 1.3293 1.3268 1.3098 1.3577 1.2680 1.3206 1.2565
    MSE 0.0612 0.0565 0.0550 0.0668 0.0494 0.0561 0.0527
    90 75 I Avg 1.3289 1.3289 1.3140 1.3521 1.2799 1.3228 1.2707
    MSE 0.0471 0.0471 0.0426 0.0502 0.0387 0.0434 0.0406
    II Avg 1.3182 1.3242 1.3010 1.3417 1.2706 1.3129 1.2613
    MSE 0.0538 0.0489 0.0485 0.0511 0.0438 0.0476 0.0469
    III Avg 1.3086 1.3106 1.2970 1.3339 1.2639 1.3057 1.2545
    MSE 0.0527 0.0496 0.0491 0.0536 0.0441 0.0477 0.0465
    90 IV Avg 1.3145 1.3156 1.3023 1.3327 1.2746 1.3096 1.2668
    MSE 0.0369 0.0336 0.0343 0.0387 0.0323 0.0347 0.0338

     | Show Table
    DownLoad: CSV
    Table A4.  Average length (AL) and coverage probability (95% CP) of asymptotic confidence/Bayesian credible interval for parameters θ = 0.4, β=1.5, λ=1.3 under different schemes.
    (k1,k2) n1=n2 m1=m2 Parameter Scheme MLE MCMC
    NR EM
    AL 95% CP AL 95% CP AL 95% CP
    (1, 2) 40 35 θ I 0.3881 0.970 0.3296 0.950 0.3033 0.957
    II 0.3881 0.980 0.3564 0.960 0.3115 0.972
    III 0.4325 0.975 0.3829 0.955 0.3197 0.972
    β I 3.0353 0.962 1.7623 0.927 1.6737 0.985
    II 3.1734 0.942 2.0625 0.945 1.6534 0.975
    III 3.6831 0.933 2.3495 0.962 1.8102 0.951
    λ I 1.4016 0.945 1.2884 0.937 1.2627 0.955
    II 1.4409 0.965 1.3230 0.953 1.2920 0.965
    III 1.4726 0.935 1.3240 0.925 1.2852 0.938
    40 θ IV 0.3418 0.970 0.3213 0.957 0.2856 0.963
    β 2.9117 0.968 2.0727 0.941 1.6887 0.983
    λ 1.3370 0.940 1.2403 0.928 1.1972 0.938
    60 40 θ I 0.344 0.955 0.2996 0.937 0.2811 0.943
    II 0.3727 0.970 0.3264 0.948 0.2997 0.955
    III 0.5527 0.959 0.3952 0.954 0.3325 0.962
    β I 2.8106 0.962 1.6974 0.952 1.6076 0.950
    II 2.6048 0.957 1.5544 0.917 1.5326 0.985
    III 4.1583 0.955 1.9933 0.925 1.8020 0.975
    λ I 1.3396 0.938 1.2094 0.945 1.2025 0.948
    II 1.3036 0.942 1.2028 0.948 1.2010 0.943
    III 1.4145 0.939 1.2123 0.943 1.2134 0.955
    50 θ I 0.2924 0.965 0.2648 0.947 0.2512 0.955
    II 0.3071 0.955 0.2832 0.945 0.2618 0.965
    III 0.3458 0.977 0.3059 0.957 0.2766 0.972
    β I 2.3854 0.953 1.6276 0.922 1.5183 0.967
    II 2.3619 0.947 1.5422 0.941 1.5310 0.977
    III 2.7614 0.944 1.7793 0.945 1.6787 0.985
    λ I 1.1633 0.932 1.0710 0.945 1.0857 0.945
    II 1.1844 0.949 1.0965 0.948 1.1030 0.937
    III 1.2236 0.954 1.0982 0.955 1.1070 0.950
    60 θ IV 0.2669 0.960 0.2528 0.945 0.2362 0.967
    β 2.1354 0.955 1.5624 0.939 1.4505 0.960
    λ 1.0524 0.952 0.9911 0.955 0.9899 0.957
    90 75 θ I 0.2349 0.948 0.2128 0.930 0.2105 0.950
    II 0.2462 0.971 0.2277 0.957 0.2196 0.967
    III 0.2682 0.960 0.2409 0.935 0.2254 0.955
    β I 1.8458 0.963 1.2457 0.965 1.2405 0.962
    II 1.8157 0.967 1.2135 0.956 1.2080 0.972
    III 2.0932 0.955 1.4069 0.942 1.3953 0.965
    λ I 0.9284 0.950 0.8610 0.953 0.8587 0.947
    II 0.9334 0.945 0.8711 0.963 0.8525 0.942
    III 0.9531 0.952 0.8737 0.948 0.8861 0.943
    90 θ IV 0.2094 0.945 0.1997 0.940 0.1916 0.952
    β 1.6631 0.953 1.2027 0.975 1.1903 0.967
    λ 0.8432 0.973 0.8003 0.970 0.8070 0.965
    (2,1) 40 35 θ I 0.4657 0.965 0.3653 0.905 0.3376 0.965
    II 0.4830 0.968 0.3919 0.940 0.3489 0.975
    III 0.5579 0.965 0.4252 0.920 0.3640 0.973
    40 IV 0.4131 0.965 0.3549 0.935 0.3182 0.955
    35 β I 2.9150 0.952 1.7654 0.939 1.6027 0.980
    II 3.0140 0.953 2.0554 0.967 1.7105 0.967
    III 3.4296 0.954 2.2692 0.967 1.7855 0.962
    IV 2.6067 0.957 1.9791 0.925 1.6166 0.962
    35 λ I 1.2645 0.922 1.2471 0.918 1.2162 0.927
    II 1.2625 0.947 1.2502 0.950 1.2287 0.962
    III 1.2594 0.928 1.2414 0.925 1.2052 0.935
    40 IV 1.1831 0.93 1.1723 0.932 1.1446 0.934
    60 40 θ I 0.4253 0.962 0.308 0.924 0.3124 0.937
    II 0.4601 0.958 0.3441 0.930 0.3377 0.972
    III 0.8594 0.965 0.4461 0.900 0.3884 0.985
    β I 2.5842 0.962 1.1519 0.929 1.4867 0.9625
    II 2.5503 0.950 1.5732 0.925 1.5298 0.980
    III 5.0418 0.942 2.0217 0.935 1.7460 0.997
    λ I 1.1723 0.933 1.1630 0.935 1.1430 0.955
    II 1.1798 0.935 1.1724 0.945 1.1460 0.963
    III 1.1909 0.945 1.1693 0.953 1.1382 0.937
    60 50 θ I 0.3655 0.958 0.2892 0.895 0.2897 0.957
    II 0.3694 0.960 0.3079 0.940 0.2939 0.967
    III 0.4519 0.971 0.3477 0.903 0.3218 0.960
    β I 2.2513 0.950 1.5107 0.932 1.4118 0.962
    II 2.2082 0.955 1.4925 0.938 1.4733 0.987
    III 2.7040 0.940 1.8107 0.942 1.6039 0.975
    λ I 1.0557 0.922 1.0435 0.920 1.0265 0.923
    II 1.0609 0.945 1.0515 0.943 1.0298 0.952
    III 1.0522 0.927 1.0401 0.934 1.0175 0.943
    60 60 θ IV 0.3124 0.960 0.2753 0.955 0.2619 0.962
    β 1.9779 0.957 1.4505 0.919 1.4023 0.977
    λ 0.9671 0.950 0.9585 0.963 0.9430 0.958
    90 75 θ I 0.2808 0.958 0.2279 0.958 0.2188 0.953
    II 0.3010 0.947 0.2505 0.932 0.2421 0.970
    III 0.3555 0.962 0.2785 0.930 0.2530 0.955
    β I 1.7621 0.940 1.3538 0.967 1.2404 0.935
    II 1.7704 0.961 1.3662 0.963 1.2725 0.956
    III 2.0689 0.943 1.4208 0.948 1.4003 0.975
    λ I 0.8634 0.960 0.8564 0.957 0.8483 0.970
    II 0.8565 0.933 0.8518 0.935 0.8419 0.943
    III 0.8587 0.933 0.8492 0.950 0.8370 0.952
    90 θ IV 0.2558 0.967 0.2260 0.960 0.2128 0.967
    β 1.5989 0.965 1.2540 0.935 1.2195 0.970
    λ 0.7798 0.950 0.7759 0.961 0.7658 0.950

     | Show Table
    DownLoad: CSV


    [1] U. Kamps, A concept of generalized order statistics, J. Statist. Plan. Inf., 48 (1995), 1–23. https://doi.org/10.1016/0378-3758(94)00147-N doi: 10.1016/0378-3758(94)00147-N
    [2] C. A. Charalambides, Discrete q-distributions on Bernoulli trials with a geometrically varying success probability, J. Statist. Plann. Inf., 140 (2010), 2355–2383. https://doi.org/10.1016/j.jspi.2010.03.024 doi: 10.1016/j.jspi.2010.03.024
    [3] R. Diaz, E. Pariguan, On the Gaussian q-distribution, J. Math. Anal. Appl., 358 (2009), 1–9. https://doi.org/10.1016/j.jmaa.2009.04.046 doi: 10.1016/j.jmaa.2009.04.046
    [4] R. Diaz, C. Ortiz, E. Pariguan, On the k-gamma q-distribution, Cent. Eur. J. Math., 8 (2010), 448–458. https://doi.org/10.2478/s11533-010-0029-0 doi: 10.2478/s11533-010-0029-0
    [5] X. Jia, S. Nadarajah, B. Guo, Inference on q-Weibull parameters, Statist. Papers, 61 (2020), 575–593. http://doi.org/10.1007/s00362-017-0951-3 doi: 10.1007/s00362-017-0951-3
    [6] B. Singh, R. U. Khan, M. A. Khan, Characterizations of q-Weibull distribution based on generalized order statistics, J. Statist. Manag. Sys., 22 (2019), 1573–1595. http://doi.org/10.1080/09720510.2019.1643554 doi: 10.1080/09720510.2019.1643554
    [7] X. Jia, Reliability analysis for q-Weibull distribution with multiply type-Ⅰ censored data, Qual. Reliab. Eng. Int., 37 (2021), 2790–2817. http://doi.org/10.1002/qre.2890 doi: 10.1002/qre.2890
    [8] F. Lad, G. Sanflippo, G. Agro, Extropy: complementary dual of entropy, Statist. Sci., 30 (2015), 40–58. http://doi.org/10.1214/14-sts430 doi: 10.1214/14-sts430
    [9] I. A. Husseiny, A. H. Syam, The extropy of concomitants of generalized order statistics from Huang-Kotz-Morgenstern bivariate distribution, J. Math., 2022 (2022), 6385998. http://dx.doi.org/10.1155/2022/6385998 doi: 10.1155/2022/6385998
    [10] S. Bansal, N. Gupta, Weighted extropies and past extropy of order statistics and k-record values, Commun. Statist. Theory Meth., 51 (2022), 6091–6108. http://dx.doi.org/10.1080/03610926.2020.1853773 doi: 10.1080/03610926.2020.1853773
    [11] S. Picoli, R. S. Mendes, L. C. Malacarne, q-exponential, Weibull, and q-Weibull distributions: an empirical analysis, Phys. A, 324 (2003), 678–688. http://doi.org/10.1016/s0378-4371(03)00071-2 doi: 10.1016/s0378-4371(03)00071-2
    [12] Z. A. Aboeleneen, Inference for Weibull distribution under generalized order statistics, Math. Comput. Sim., 81 (2010), 26–36. http://dx.doi.org/10.1016/j.matcom.2010.06.013 doi: 10.1016/j.matcom.2010.06.013
    [13] A. A. Jafari, H. Zakerzadeh, Inference on the parameters of the Weibull distribution using records, SORT, 39 (2015), 3–18.
    [14] P. H. Garthwaite, J. B. Kadane, A. OHagan, Statistical methods for eliciting probability distributions, J. Am. Stat. Assoc., 100 (2005), 680–701. http://dx.doi.org/10.1198/016214505000000105 doi: 10.1198/016214505000000105
    [15] K. M. Hamdia, X. Zhuang, P. He, T. Rabczuk, Fracture toughness of polymeric particle nanocomposites: Evaluation of models performance using Bayesian method, Composites Sci. Tech., 126 (2016), 122–129. http://dx.doi.org/10.1016/j.compscitech.2016.02.012 doi: 10.1016/j.compscitech.2016.02.012
    [16] X. Jia, D. Wang, P. Jiang, B. Guo, Inference on the reliability of Weibull distribution with multiply type-Ⅰ censored data, Reliab. Eng. Syst. Saf., 150 (2016), 171–181. http://doi.org/10.1016/j.ress.2016.01.025 doi: 10.1016/j.ress.2016.01.025
    [17] M. Xu, E. L. Droguett, I. D. Lins, M. das Chagas Moura, On the q-Weibull distribution for reliability applications: An adaptive hybrid artificial bee colony algorithm for parameter estimation, Reliab. Eng. Syst. Saf., 158 (2017), 93–105. http://doi.org/10.1016/j.ress.2016.10.012 doi: 10.1016/j.ress.2016.10.012
    [18] R. B. Nelsen, An Introduction to Copulas, New York: Springer-Verlag, 2006.
    [19] A. Sklar, Random variables, joint distribution functions, and copulas, Kybernetika, 9 (1973), 449–460.
    [20] E. J. Gumbel, Bivariate exponential distributions, J. Am. Stat. Assoc., 55 (1960), 698–707.
    [21] M. A. Alawady, H. M. Barakat, M. A. Abd Elgawad, Concomitants of generalized order statistics from bivariate Cambanis family of distributions under a general setting, Bull. Malays. Math. Sci. Soc., 44 (2021), 3129–3159. http://dx.doi.org/10.1007/s40840-021-01102-1 doi: 10.1007/s40840-021-01102-1
    [22] M. A. Alawady, H. M. Barakat, S. Xiong, M. A. Abd Elgawad, Concomitants of generalized order statistics from iterated Farlie-Gumbel-Morgenstern type bivariate distribution, Commun. Statist. Theory Meth., 51 (2022), 5488–5504. http://dx.doi.org/10.1080/03610926.2020.1842452 doi: 10.1080/03610926.2020.1842452
    [23] S. P. Arun, C. Chesneau, R. Maya, M. R. Irshad, Farlie-Gumbel-Morgenstern bivariate moment exponential distribution and its inferences based on concomitants of order statistics, Stats, 6 (2023), 253–267. http://dx.doi.org/10.3390/stats6010015 doi: 10.3390/stats6010015
    [24] H. M. Barakat, M. A. Alawady, I. A. Husseiny, G. M. Mansour, Sarmanov family of bivariate distributions: statistical properties, concomitants of order statistics, and information measures, Bull. Malays. Math. Sci. Soc., 45 (2022), 49–83. http://dx.doi.org/10.1007/s40840-022-01241-z doi: 10.1007/s40840-022-01241-z
    [25] H. M. Barakat, E. M. Nigm, M. A. Alawady, I. A. Husseiny, Concomitants of order statistics and record values from the generalization of FGM bivariate-generalized exponential distribution, J. Statist. Theory Appl., 18 (2019), 309–322. http://dx.doi.org/10.2991/jsta.d.190822.001 doi: 10.2991/jsta.d.190822.001
    [26] H. M. Barakat, E. M. Nigm, M. A. Alawady, I. A. Husseiny, Concomitants of order statistics and record values from the iterated FGM type bivariate-generalized exponential distribution, REVSTAT Statist. J., 19 (2021), 291–307. http://dx.doi.org/10.2298/fil1809313b doi: 10.2298/fil1809313b
    [27] S. Cambanis, Some properties and generalizations of multivariate Eyraud-Gumbel-Morgenstern distributions, J. Mul. Anal., 7 (1977), 551–559. http://dx.doi.org/10.1016/0047-259x(77)90066-5 doi: 10.1016/0047-259x(77)90066-5
    [28] I. A. Husseiny, H. M. Barakat, G. M. Mansour, M. A. Alawady, Information measures in record and their concomitants arising from Sarmanov family of bivariate distributions, J. Comput. Appl. Math., 408 (2022), 114120. http://doi.org/10.1016/j.cam.2022.114120 doi: 10.1016/j.cam.2022.114120
    [29] I. A. Husseiny, M. A. Alawady, H. M. Barakat, M. A. Abd Elgawad, Information measures for order statistics and their concomitants from Cambanis bivariate family, Commun. Statist. Theory Meth., 53 (2024), 865–881. http://doi.org/10.1080/03610926.2022.2093909 doi: 10.1080/03610926.2022.2093909
    [30] J. Scaria, B. Thomas, Second order concomitants from the Morgenstern family of distributions, J. Appl. Statist. Sci., 21 (2014), 63–76.
    [31] W. Schucany, W. C. Parr, J. E. Boyer, Correlation structure in Farlie-Gumbel-Morgenstern distributions, Biometrika, 65 (1978), 650–653. http://doi.org/10.1093/biomet/65.3.650 doi: 10.1093/biomet/65.3.650
    [32] R. A. Attwa, T. Radwan, E. O. Abo Zaid, Bivariate q-extended Weibull Morgenstern family and correlation coefficient formulas for some of its sub-models, AIMS Math., 8 (2023), 25325–25342. http://dx.doi.org/10.3934/math.20231292 doi: 10.3934/math.20231292
    [33] M. Gurvich, A. Dibenedetto, S. Ranade, A new statistical distribution for characterizing the random strength of brittle materials, J. Mater. Sci., 32 (1997), 2559–2564.
    [34] H. A. David, Concomitants of order statistics, Bull. Int. Statist. Inst., 45 (1973), 295–300.
    [35] I. Bairamov, S. Kotz, M. Becki, New generalized Farlie-Gumbel-Morgenstern distributions and concomitants of order statistics, J. Appl. Statist., 28 (2001), 521–536. http://dx.doi.org/10.1080/02664760120047861 doi: 10.1080/02664760120047861
    [36] M. I. Beg, M. Ahsanullah, Concomitants of generalized order statistics from Farlie-Gumbel-Morgenstern distributions, Statist. Methodol., 5 (2008), 1–20. http://dx.doi.org/10.1016/j.stamet.2007.04.001 doi: 10.1016/j.stamet.2007.04.001
    [37] F. Domma, S. Giordano, Concomitants of m-generalized order statistics from generalized Farlie-Gumbel-Morgenstern distribution family, J. Comput. Appl. Math., 294 (2016), 413–435. http://dx.doi.org/10.1016/j.cam.2015.08.022 doi: 10.1016/j.cam.2015.08.022
    [38] S. Eryilmaz, On an application of concomitants of order statistics, Commun. Statist. Theory Meth., 45 (2016), 5628–5636. http://dx.doi.org/10.1080/03610926.2014.948201 doi: 10.1080/03610926.2014.948201
    [39] J. Scaria, N. U. Nair, Distribution of extremes of rth concomitant from the Morgenstern family, Statist. Papers, 49 (2008), 109–119. http://doi.org/10.1007/s00362-006-0365-0 doi: 10.1007/s00362-006-0365-0
    [40] S. Tahmasebi, A. A. Jafari, Concomitants of order statistics and record values from Morgenstern type bivariate-generalized exponential distribution, Bull. Malays. Math. Sci. Soc., 38 (2015), 1411–1423. http://doi.org/10.1007/s40840-014-0087-8 doi: 10.1007/s40840-014-0087-8
    [41] H. S. Klakattawi, W. H. Aljuhani, L. A. Baharith, Alpha power exponentiated new Weibull-Pareto distribution: Its properties and applications, Pakistan J. Statist. Oper. Res., 18 (2022), 703–720. http://doi.org/10.18187/pjsor.v18i3.3937 doi: 10.18187/pjsor.v18i3.3937
    [42] E. T. Lee, J. Wang, Statistical Methods for Survival Data Analysis, New york: John Wiley & Sons, 2003.
  • This article has been cited by:

    1. Abd Ulazeez Alkouri, Eman A. Abuhijleh, Eman Almuhur, Ghada Alafifi, Sana Abu-Ghurra, Subgroups and Homomorphism Structures of Complex Pythagorean Fuzzy Sets, 2024, 23, 2224-2880, 614, 10.37394/23206.2024.23.65
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1303) PDF downloads(68) Cited by(3)

Figures and Tables

Figures(10)  /  Tables(7)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog